Cerebras WSE-3: Third Era Superchip for AI

Sunnyvale, Calif., AI supercomputer agency Cerebras says its subsequent era of waferscale AI chips can do double the efficiency of the earlier era whereas consuming the identical quantity of energy. The Wafer Scale Engine 3 (WSE-3) incorporates 4 trillion transistors, a greater than 50 % improve over the earlier era because of using newer chipmaking expertise. The corporate says it would use the WSE-3 in a brand new era of AI computer systems, which at the moment are being put in in a datacenter in Dallas to kind a supercomputer able to 8 exaflops (8 billion billion floating level operations per second). Individually, Cerebras has entered right into a joint improvement settlement with Qualcomm that goals to spice up a metric of worth and efficiency for AI inference 10-fold.

The corporate says the CS-3 can practice neural community fashions as much as 24-trillion parameters in measurement, greater than 10 occasions the scale of at present’s largest LLMs.

With WSE-3, Cerebras can maintain its declare to producing the biggest single chip on the earth. Sq.-shaped with 21.5 centimeters to a facet, it makes use of almost a whole 300-millimeter wafer of silicon to make one chip. Chipmaking tools is often restricted to producing silicon dies of not more than about 800 sq. millimeters. Chipmakers have begun to flee that restrict by utilizing 3D integration and different superior packaging expertise3D integration and different superior packaging expertise to mix a number of dies. However even in these methods, the transistor depend is within the tens of billions.

As regular, such a big chip comes with some mind-blowing superlatives.

You possibly can see the impact of Moore’s Regulation within the succession of WSE chips. The primary, debuting in 2019, was made utilizing TSMC’s 16-nanometer tech. For WSE-2, which arrived in 2021, Cerebras moved on to TSMC’s 7-nm course of. WSE-3 is constructed with the foundry big’s 5-nm tech.

The variety of transistors has greater than tripled since that first megachip. In the meantime, what they’re getting used for has additionally modified. For instance, the variety of AI cores on the chip has considerably leveled off, as has the quantity of reminiscence and the interior bandwidth. However, the advance in efficiency by way of floating-point operations per second (flops) has outpaced all different measures.

CS-3 and the Condor Galaxy 3

The pc constructed across the new AI chip, the CS-3, is designed to coach new generations of big massive language fashions, 10 occasions bigger than OpenAI’s GPT-4 and Google’s Gemini. The corporate says the CS-3 can practice neural community fashions as much as 24-trillion parameters in measurement, greater than 10 occasions the scale of at present’s largest LLMs, with out resorting to a set of software program methods wanted by different computer systems. Based on Cerebras, which means the software program wanted to coach a one-trillion parameter mannequin on the CS-3 is as simple as coaching a one billion parameter mannequin on GPUs.

As many as 2,048 methods will be mixed, a configuration that might chew by way of coaching the favored LLM Llama 70B from scratch in simply in the future. Nothing fairly that large is within the works, although, the corporate says. The primary CS-3-based supercomputer, Condor Galaxy 3 in Dallas, can be made up of 64 CS-3s. As with its CS-2-based sibling methods, Abu Dhabi’s G42 owns the system. Along with Condor Galaxy 1 and a couple of, that makes a community of 16 exaflops.

“The prevailing Condor Galaxy community has skilled a number of the main open-source fashions within the trade, with tens of hundreds of downloads,” mentioned Kiril Evtimov, group CTO of G42 in a press launch. “By doubling the capability to 16 exaflops, we sit up for seeing the following wave of innovation Condor Galaxy supercomputers can allow.”

A Deal With Qualcomm

Whereas Cerebras computer systems are constructed for coaching, Cerebras CEO Andrew Feldman says it’s inference, the execution of neural community fashions, that’s the actual restrict to AI’s adoption. Based on Cerebras estimates, if each particular person on the planet used ChatGPT, it could price US $1 trillion yearly—to not point out an amazing quantity of fossil-fueled power. (Working prices are proportional to the scale of neural community mannequin and the variety of customers.)

So Cerebras and Qualcomm have shaped a partnership with the objective of bringing the price of inference down by an element of 10. Cerebras says their answer will contain making use of neural networks methods comparable to weight knowledge compression and sparsity—the pruning of unneeded connections. The Cerebras-trained networks would then run effectively on Qualcomm’s new inference chip, the AI 100 Extremely, the corporate says.